EVA: Efficient Reinforcement Learning for End-to-End Video Agent - Explained Simply | ArXiv Explained